o Kdump broke in 2.6.19-rc4-mm1 on i386 due to LAPIC timer interrupt handling
changes. (clock events)
o Problems happens because a pending LAPIC timer interrupt is received
as soon as kernel enables interrupts first time during boot but
associated clock event device has not been initialized yet and
kernel crashes.
o One can receive a pending LAPIC timer interrupt from previous kernel's
context (as in kdump).
o A pending LAPIC timer interrupt is delivedred to CPU in second kernel as
soon as interrupts are enabled. Already idt vector has been set to handle
LAPIC timer interrupts but actually LPAIC is setup in setup_local_APIC()
much later. setup_local_APIC() also initializes LAPIC timer and in turn
related clock_event_device. But here interrupts comes really early and LAPIC
has not been set, associated clock_event_device structure has not been
initialized hence kernel crashes in smp_apic_timer_interrupt().
o One of the possible solution is that register the LAPIC timer vector late
after LAPIC has been initialized. But this would only solve the problem
for boot cpu and not other cpus.
o In this patch, I check for clock_event_device event handler and if
handler is NULL, that interrupt is ignored as spurious one. Not the best
way to handle the situation as it does not count for bugs in
clock_event_device creation and registration logic, and also introduces
un-necessary check in fastpath. Pleaes suggest if there is a better way
to handle this.
Signed-off-by: Vivek Goyal <[email protected]>
---
arch/i386/kernel/apic.c | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
diff -puN arch/i386/kernel/apic.c~debug1-patch arch/i386/kernel/apic.c
--- linux-2.6.19-rc4-reloc/arch/i386/kernel/apic.c~debug1-patch 2006-11-01 15:30:03.000000000 -0500
+++ linux-2.6.19-rc4-reloc-root/arch/i386/kernel/apic.c 2006-11-01 17:19:06.000000000 -0500
@@ -1322,6 +1322,23 @@ fastcall void smp_apic_timer_interrupt(s
int cpu = smp_processor_id();
struct clock_event_device *evt = &per_cpu(lapic_events, cpu);
+ /* Normally we should not be here till LAPIC has been initialized
+ * but in some cases like kdump, its possible that there is a
+ * pending LAPIC timer interrupt from previous kernel's context
+ * and is delivered in new kernel the moment interrupts are enabled.
+ *
+ * Interrupts are enabled early and LAPIC is setup much later, hence
+ * its possible that when we get here evt->event_handler is NULL.
+ * Check for event_handler being NULL and discard the interrupt
+ * as spurious.
+ */
+
+ if (!evt->event_handler) {
+ printk("Spurious LAPIC timer interrupt on cpu %d\n", cpu);
+ set_irq_regs(old_regs);
+ return;
+ }
+
/*
* the NMI deadlock-detector uses this.
*/
_